STARWEST 2026 - Industry Insights

Coding agents now write a large and growing share of production code, and they'll happily write the tests for it too. But an agent grading its own homework isn't verification. This session looks at why the author cannot be the verifier, what breaks when AI checks its own work (drawing on recent research into agent overconfidence and reward hacking), and a practical framework for independent verification that keeps quality ahead of AI-speed development. We'll close on where human testers decisively win, and how QA shifts from writing tests to owning quality strategy.

Does the tester survive AI? Wrong Question. Agents now write both code and tests, and the industry's instinct-putting a human downstream to grade output-doesn't scale. It also wastes the quality engineer's judgment in the least useful spot possible.

The real answer: teach agents your standard once, so every session inherits it automatically. That's Quality Experience Lay (QEL).

Financial Services. Legal. Higher Education. Pharma. Four very different industries, but all facing the same fundamental shift: what happens when AI agents become part of the software delivery and quality engineering team?

Most quality organizations have capable people and tools that (usually) get the job done. The coordination between them is where releases actually slow down. QA & Testing activities run across a variety of tools and reports; a mobile tool, a separate API tool, a device farm on its own contract, performance owned by another team, and results reconciled by hand. Agentic delivery raises the price of those gaps. When agents generate tests, repair them, and execute continuously, authoring stops being the constraint and coordination takes its place.

The history of enterprise software follows a familiar pattern.

Linux. Git. Docker. Kubernetes. Terraform. Each began as a powerful open source project embraced by developers before eventually becoming foundational enterprise technology. Along the way, every one encountered the same challenge: the growing gap between what individual developers need and what organizations require to operate those technologies securely, reliably, and at scale.

Today, Playwright is following the same path.

More and more companies are cutting costs and doing layoffs, and QA is often among the first to be cut. Yet it is precisely us QA folks who can lead cost cutting initiatives while improving quality and saving some time and effort too!In this talk, I suggest ways to gather numbers for the internal and external failure costs the company is already paying.

As software development pivots to AI-driven agents and autonomous workflows, quality assurance stakes have never been higher. Inflectra CEO Adam Sandman will demonstrate the company's latest AI innovations.

Confidently wrong AI has a price tag, and in 2026, it's a line item. Your AI agent will never tell you it's wrong — it doesn't know it's wrong. It ships code and answers customers with total confidence, whether it's right or has just invented a refund policy that doesn't exist. Confidence isn't correctness, and an agent can't grade its own work. Failures look like green checkmarks — until a customer or regulator finds one. 99% of organizations deploying AI report losses, averaging $4.4M each. The fix: an independent layer that verifies agents before they reach customers, not after.

Developers don't write most of their code from scratch anymore, they generate it. That's already changed the pace of software delivery, and it's now landing directly on QA: more changes, more code, more required coverage, without bigger teams.
This session looks at what that means for testing in practice. We'll walk through a workflow where AI does real work: turning a requirement into a starting point instead of a blank page, deciding what's worth automating, and scoping regression to what actually changed.

You already know context is everything for AI, and MCP has become the go-to way to feed it. But dumping raw data at your AI isn't a strategy. Real value comes from the tools you build around that context, connecting requirements, tests, and defects to capture true product intent. Xray's MCP uses AI and semantic search to turn your test data into meaningful context fueling every SDLC phase: plan, code, validate, release. Come observe an end-to-end agentic workflow pairing Xray's MCP and skills with Jira Assignable Agents to move work from idea to release with QE expertise baked in.

AI is moving beyond assistance toward systems that can reason, pursue goals, use tools, and collaborate with other agents. What does this shift mean for Quality Engineering? This session provides a practical path from traditional test automation to Agentic QE.

AI agents are moving beyond simple test case generation into active participants across the entire Software Testing Life Cycle. This session walks through how autonomous and semi-autonomous agents can support each STLC phase, turning requirements into testable scenarios, generating and prioritizing test cases, executing and self-healing automated suites, triaging failures, and synthesizing coverage reports, while highlighting where human judgment still has to stay in the loop.